DocTable AllocateDocTable(void) {
DocTableRecord *dt = (DocTableRecord *) malloc(sizeof(DocTableRecord));
Verify333(dt != NULL);
dt->docid_to_docname = AllocateHashTable(1024);
dt->docname_to_docid = AllocateHashTable(1024);
dt->max_id = 0;
Verify333(dt->docid_to_docname != NULL);
Verify333(dt->docname_to_docid != NULL);
return dt;
}
static void TestFindOrInsert() {
struct HashTable* ht;
int i;
int iterations = 1000000;
int range = 30; /* random number between 1 and 30 */
ht = AllocateHashTable(4, 0); /* value is 4 bytes, 0: don't copy keys */
/* We'll test how good rand() is as a random number generator */
for (i = 0; i < iterations; ++i) {
int key = rand() % range;
HTItem* bck = HashFindOrInsert(ht, key, 0); /* initialize to 0 */
bck->data++; /* found one more of them */
}
for (i = 0; i < range; ++i) {
HTItem* bck = HashFind(ht, i);
if (bck) {
printf("%3d: %d\n", bck->key, bck->data);
} else {
printf("%3d: 0\n", i);
}
}
FreeHashTable(ht);
}
static void ResizeHashtable(HashTable ht) {
// Resize if the load factor is > 3.
if (ht->num_elements < 3 * ht->num_buckets)
return;
// This is the resize case. Allocate a new hashtable,
// iterate over the old hashtable, do the surgery on
// the old hashtable record and free up the new hashtable
// record.
HashTable newht = AllocateHashTable(ht->num_buckets * 9);
// Give up if out of memory.
if (newht == NULL)
return;
// Loop through the old ht with an iterator,
// inserting into the new HT.
HTIter it = HashTableMakeIterator(ht);
if (it == NULL) {
// Give up if out of memory.
FreeHashTable(newht, &HTNullFree);
return;
}
while (!HTIteratorPastEnd(it)) {
HTKeyValue item, dummy;
Verify333(HTIteratorGet(it, &item) == 1);
if (InsertHashTable(newht, item, &dummy) != 1) {
// failure, free up everything, return.
HTIteratorFree(it);
FreeHashTable(newht, &HTNullFree);
return;
}
HTIteratorNext(it);
}
// Worked! Free the iterator.
HTIteratorFree(it);
// Sneaky: swap the structures, then free the new table,
// and we're done.
{
HashTableRecord tmp;
tmp = *ht;
*ht = *newht;
*newht = tmp;
FreeHashTable(newht, &HTNullFree);
}
return;
}
HashTable BuildWordHT(char *filename) {
char *filecontent;
HashTable tab;
HWSize_t filelen, i;
if (filename == NULL)
return NULL;
// STEP 6.
// Use ReadFile() to slurp in the file contents. If the
// file turns out to be empty (i.e., its length is 0),
// or you couldn't read the file at all, return NULL to indicate
// failure.
filecontent = ReadFile(filename, &filelen);
if (filecontent == NULL || filelen == 0)
return NULL;
// Verify that the file contains only ASCII text. We won't try to index any
// files that contain non-ASCII text; unfortunately, this means we aren't
// Unicode friendly.
for (i = 0; i < filelen; i++) {
if ((filecontent[i] == '\0') ||
((unsigned char) filecontent[i] > ASCII_UPPER_BOUND)) {
free(filecontent);
return NULL;
}
}
// Great! Let's split the file up into words. We'll allocate the hash
// table that will store the WordPositions structures associated with each
// word. Since our hash table dynamically grows, we'll start with a small
// number of buckets.
tab = AllocateHashTable(64);
// Loop through the file, splitting it into words and inserting a record for
// each word.
LoopAndInsert(tab, filecontent);
// If we found no words, return NULL instead of a
// zero-sized hashtable.
if (NumElementsInHashTable(tab) == 0) {
FreeHashTable(tab, &WordHTFree);
tab = NULL;
}
// Now that we've finished parsing the document, we can free up the
// filecontent buffer and return our built-up table.
free(filecontent);
filecontent = NULL;
return tab;
}
static void TestInsert() {
struct HashTable* ht;
HTItem* bck;
ht = AllocateHashTable(1, 0); /* value is 1 byte, 0: don't copy keys */
HashInsert(ht, PTR_KEY(ht, "January"), 31); /* 0: don't overwrite old val */
bck = HashInsert(ht, PTR_KEY(ht, "February"), 28);
bck = HashInsert(ht, PTR_KEY(ht, "March"), 31);
bck = HashFind(ht, PTR_KEY(ht, "February"));
assert(bck);
assert(bck->data == 28);
FreeHashTable(ht);
}
gar_list* gar_index(void* gar, size_t length)
{
gar_list* list = malloc(sizeof(gar_list));
size_t size;
char name[100];
unsigned long i;
name[99] = '\0';
list->gar = gar;
list->length = length;
list->ht = AllocateHashTable(0, 1);
for (; length >= 512; gar += 512, length -= 512)
if (!*(char*)gar)
return list;
else
{
size = strtol(gar + 124, NULL, 8);
if (((char*)gar)[156] == '0' || ((char*)gar)[156] == '\0')
{
for (i = 5; i < 99; i ++)
switch ((name[i - 5] = ((char*)gar)[i]))
{
case '/':
name[i - 5] = '.';
break;
case ' ':
name[i - 5] = '\0';
case '\0':
goto copied;
}
copied:
HashInsert(list->ht, PTR_KEY(list->ht, name), (ulong)gar + 512);
}
size = size ? ((size - 1) / 512 + 1) * 512 : 0;
if (length < size)
break;
gar += size;
length -= size;
}
FreeHashTable(list->ht);
free(list);
return NULL;
}
int main (int argc, char *argv[])
{
paper_rec_t DedupeRecord;
dd_uint64_t Unique_CRID; /* Unique CR_ID = (C_ID << 16) | CR_ID */
long victim_index = 0, cache_size, window_size, bloom_filter_size;
long i, j=0, temp_index;
int Initial_Flag = 0, cache_algorithm;
dd_uint8_t *sha1_value=NULL;
int nLen = 0;
long max_counter=0;
HTItem *chunk_item, *item;
long byte_len, temp, offset, ver, temp1; /* to read a trace file */
unsigned long hash1, hash2;
/* Heap Data structure variables */
Cache_Memory Dataitem;
std::vector<Cache_Memory>::iterator itr;
unsigned long writeCounter = 0;
unsigned long access_counter;
long file_position;
FILE *fp1, *fp;
size_t keySize=0,iCnt;
clock_t start = clock();
time_t begin,end;
time(&begin);
if (argc < 5) {
/* 0 1 2 3 4 */
fprintf(stderr, "usage: %s <Cache Size> <Window Size> <Cache Algorithm (0, 1, 2)> <Trace File>\n", argv[0]);
fprintf(stderr, " - Cache Size: Dedupe read cache size in terms of # of data chunk (e.g. 500 chunks = 4MB (500*8KB))\n");
fprintf(stderr, " - Window Size: Future sliding window size in terms of TIMES of cache size.\n");
fprintf(stderr, " - Cache Algorithm: 0 (Belady MIN), 1 (Belady MIN with a future window), 2 (Lookahead read cache)\n");
fprintf(stderr, " - Trace File: Trace file name with path\n");
exit(1);
}
cache_size = atol(argv[1]);
assert(cache_size > 0); /* cache size must be positive */
window_size = atol(argv[2]);
assert(window_size > 0); /* window size must be positive */
cache_algorithm = atoi(argv[3]);
assert((cache_algorithm == 0)||(cache_algorithm == 1)||(cache_algorithm == 2)); /* there are only three selections */
bloom_filter_size = cache_size*2; //No. of Hash functions for BF is 2
bloom_filter = (long *)malloc(sizeof(long)*bloom_filter_size);
ht_cache = AllocateHashTable(SHA1_VALUE_LENGTH, 1);
heap = newMinHeap((u32)cache_size);
if((fp1 = fopen(argv[4], "rb")) == NULL){ //for reading data one by one
DEBUG_INFO("File open error....1\n");
exit (-1);
}
if((fp = fopen(argv[4], "rb")) == NULL){ //for searching its future reference distance
DEBUG_INFO("File open error....2\n");
exit (-1);
}
long c=0, d=0;
u32 itemIndex;
keySize = sizeof(DedupeRecord.fp_bytes);
DEBUG_INFO("Record Size is: %d\n",keySize);
while (1)
{
fread(&DedupeRecord, sizeof(struct _paper_rec_t),1, fp1);
/*if(DedupeRecord.fp_bytes[0] == 0)
DedupeRecord.fp_bytes[0] = '0';*/
/*for(iCnt = 0;iCnt<sizeof(DedupeRecord.fp_bytes);++iCnt)
printf("%c",DedupeRecord.fp_bytes[iCnt]);*/
//DEBUG_INFO("Reading chunks : %ld\n", c++);
c++;
if(c%1000 == 0){
printf("Reading Chunks: %ld\n",c);
}
if (c % 10000 == 0) {
printf("Cache hit ratio: %.3f = %lu / %lu \n",
(double) (Hit_Count * 100) / (double) totalAccess ,
Hit_Count,
totalAccess);
}
if(feof(fp1))
break;
file_position = ftell(fp1);
/* initially fill the cache. During this initialization process, we do not count the cache hit ratio. */
if (Initial_Flag == 0) {
// Temporally store this current access chunk with its future reference distance in the cache
chunk_item = HashFind(ht_cache, PTR_KEY(ht_cache,DedupeRecord.fp_bytes));
//Update Bloom filter counters
hash1 = hash_djb2(DedupeRecord.fp_bytes,keySize)%bloom_filter_size;
hash2 = hash_sdbm(DedupeRecord.fp_bytes,keySize)%bloom_filter_size;
max_counter = bloom_filter[hash1]++;
if((bloom_filter[hash2]++) > max_counter)
max_counter = bloom_filter[hash2];
if(chunk_item) { //Cache Hit
itemIndex = (u32)chunk_item->data;
DEBUG_INFO("Index its updating is %ld:\n",itemIndex);
heapUpdate(heap,max_counter,itemIndex,&ht_cache);
}
else {
heapInsert(heap,DedupeRecord.fp_bytes, max_counter,&ht_cache);
//Sandeep - Insert into Heap and Heapify
cache_counter++;
}
if(cache_counter == cache_size) {
DEBUG_INFO("\n#### Cache Initialization complete~!!####\n");
Initial_Flag = 1;
//Sandeep - Construct Heap and Heapify
//fnBuildHeap(cache_heap);
#ifdef DEBUG
printf("Heap Size is: %d\n",cache_heap.size());
/*PrintHashTable(ht_cache,-1,2);
fnPrintHeap(cache_heap);*/
#endif
}
}
else { /* Once the cache is full of data initially, we start to measure the cache hit ratio from now. */
totalAccess++;
if((totalAccess % 100) == 0) {
DEBUG_INFO("[CHECK] Current Access Number: %ld\n", totalAccess);
}
Unique_CRID = (DedupeRecord.cmc_id << 16) | DedupeRecord.creg_id;
chunk_item = HashFind(ht_cache, PTR_KEY(ht_cache,DedupeRecord.fp_bytes));
if(chunk_item) { //Cache Hit
Hit_Count++;
DEBUG_INFO("Cache Hit\n");
//Update Bloom filter counters
hash1 = hash_djb2(DedupeRecord.fp_bytes,keySize)%bloom_filter_size;
hash2 = hash_sdbm(DedupeRecord.fp_bytes,keySize)%bloom_filter_size;
//DEBUG_INFO("### Returned hash values are %ld and %ld\n",bloom_filter[hash1],bloom_filter[hash2]);
max_counter = bloom_filter[hash1]++;
if((bloom_filter[hash2]++) > max_counter)
max_counter = bloom_filter[hash2];
itemIndex = (ulong)chunk_item->data;
DEBUG_INFO("Index its updating is %ld:\n",itemIndex);
assert(itemIndex>=0 && itemIndex<=cache_size);
heapUpdate(heap,max_counter,itemIndex,&ht_cache);
//Sandeep - Update heap counter val for this chunk with max_counter
//fnUpdateHeap(cache_heap, Read_Cache[(ulong)chunk_item->data],max_counter);
}
else {
heapPopMin(heap,&sha1_value,&access_counter,&ht_cache);
if(!sha1_value)
ERROR("SHA1 Value in main is NULL\n");
/*for(iCnt = 0;iCnt<sizeof(DedupeRecord.fp_bytes);++iCnt)
printf("%c",sha1_value[iCnt]);*/
//Update Bloom filter counters
hash1 = hash_djb2(sha1_value,sizeof(sha1_value))%bloom_filter_size;
hash2 = hash_sdbm(sha1_value,sizeof(sha1_value))%bloom_filter_size;
//DEBUG_INFO("### In Main before decrement %ld and %ld\n",bloom_filter[hash1],bloom_filter[hash2]);
//Decrement BF counters
bloom_filter[hash1]--;
bloom_filter[hash2]--;
free(sha1_value);
//GP - Increment the BF counters for this chunk
hash1 = hash_djb2(DedupeRecord.fp_bytes,keySize)%bloom_filter_size;
hash2 = hash_sdbm(DedupeRecord.fp_bytes,keySize)%bloom_filter_size;
//DEBUG_INFO("### Returned hash values are in main cache_miss %ld and %ld\n",bloom_filter[hash1],bloom_filter[hash2]);
max_counter = bloom_filter[hash1]++;
if((bloom_filter[hash2]++) > max_counter)
max_counter = bloom_filter[hash2];
heapInsert(heap,DedupeRecord.fp_bytes,max_counter,&ht_cache);
if(cache_algorithm == LOOKAHEAD){
/* Check if any other chunks in the current CR will appear within the future window.
If we found one, we add such chunk(s) in the cache. */
Check_Unique_CRID(fp, Unique_CRID, file_position, 0, cache_size, window_size*cache_size, bloom_filter_size);
}
}
} //else
} //while
printf("\n###################################################################\n");
printf("Cache hit ratio: %.3f = %lu / %lu \n", (double) (Hit_Count * 100) / (double) totalAccess , Hit_Count, totalAccess);
printf("Cache size: %ld, window size: %ld\n", cache_size, window_size*cache_size);
printf("Dedupe trace: %s\n", argv[4]);
printf("###################################################################\n");
fclose(fp1);
fclose(fp);
FreeHashTable(ht_cache);
deleteMinHeap(heap);
time(&end);
printf("###################################################################\n");
printf("Total time taken is %f \n",((double)clock()-start)/CLOCKS_PER_SEC);
printf("###################################################################\n");
return 0;
}
// our main function; here, we demonstrate how to use some
// of the hash table functions
int main(int argc, char **argv) {
ExampleValuePtr evp;
HashTable ht;
HTIter iter;
HTKeyValue kv, old_kv;
HTKey_t i;
// allocate a hash table with 10,000 initial buckets
ht = AllocateHashTable(10000);
Verify333(ht != NULL);
// insert 20,000 elements (load factor = 2.0)
for (i = 0; i < 20000; i++) {
evp = (ExampleValuePtr) malloc(sizeof(ExampleValue));
Verify333(evp != NULL);
evp->num = i;
// make sure HT has the right # of elements in it to start
Verify333(NumElementsInHashTable(ht) == (HWSize_t) i);
// insert a new element
kv.key = FNVHashInt64((HTValue_t)i);
kv.value = (HTValue_t)evp;
Verify333(InsertHashTable(ht, kv, &old_kv) == 1);
// make sure hash table has right # of elements post-insert
Verify333(NumElementsInHashTable(ht) == (HWSize_t) (i+1));
}
// look up a few values
Verify333(LookupHashTable(ht, FNVHashInt64((HTValue_t)100), &kv) == 1);
Verify333(kv.key == FNVHashInt64((HTValue_t)100));
Verify333(((ExampleValuePtr) kv.value)->num == 100);
Verify333(LookupHashTable(ht, FNVHashInt64((HTValue_t)18583), &kv) == 1);
Verify333(kv.key == FNVHashInt64((HTValue_t)18583));
Verify333(((ExampleValuePtr) kv.value)->num == 18583);
// make sure non-existent value cannot be found
Verify333(LookupHashTable(ht, FNVHashInt64((HTValue_t)20000), &kv) == 0);
// delete a value
Verify333(RemoveFromHashTable(ht, FNVHashInt64((HTValue_t)100), &kv) == 1);
Verify333(kv.key == FNVHashInt64((HTValue_t)100));
Verify333(((ExampleValuePtr) kv.value)->num == 100);
ExampleValueFree(kv.value); // since we malloc'ed it, we must free it
// make sure it's deleted
Verify333(LookupHashTable(ht, FNVHashInt64((HTValue_t)100), &kv) == 0);
Verify333(NumElementsInHashTable(ht) == (HWSize_t) 19999);
// loop through using an iterator
i = 0;
iter = HashTableMakeIterator(ht);
Verify333(iter != NULL);
while (HTIteratorPastEnd(iter) == 0) {
Verify333(HTIteratorGet(iter, &kv) == 1);
Verify333(kv.key != FNVHashInt64((HTValue_t)100)); // we deleted it
// advance the iterator
HTIteratorNext(iter);
i++;
}
Verify333(i == 19999);
// free the iterator
HTIteratorFree(iter);
// free the hash table
FreeHashTable(ht, &ExampleValueFree);
return 0;
}
